Mixers, Modulators and Frequency Conversion

Introduction and Context

Frequency conversion is central to virtually every radio-frequency (RF) and microwave system. Antennas, propagation channels, filters, amplifiers and data converters operate over different frequency ranges, while a communication system must translate information between those ranges. Mixers perform this translation by exploiting the nonlinear relationship between two or more signals. Modulators deliberately create a spectrum containing information-bearing sidebands, whereas demodulators recover the information. In many practical circuits the same device can serve as a mixer, modulator or demodulator depending on how its ports are driven and how its output is filtered.

A receiver may translate a weak signal from a radio frequency to an intermediate frequency (IF), where high-selectivity filters and amplifiers are convenient. A transmitter may translate a digitally modulated IF signal to its assigned RF channel. Direct-conversion architectures translate directly between RF and zero frequency, while superheterodyne architectures use one or more non-zero IF stages.

At postgraduate level, frequency conversion must be understood simultaneously in four ways: as a time-domain nonlinear operation, as a frequency-domain convolution, as a circuit design problem involving conversion gain and impedance matching, and as a system problem involving unwanted products, noise, linearity and frequency planning.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain how nonlinear devices generate sum and difference frequencies.
2. Derive the principal mixer and modulation relationships.
3. Distinguish conversion gain, conversion loss, isolation, noise figure and linearity.
4. Analyse single-balanced, double-balanced, passive, active and image-reject mixers.
5. Design a basic frequency plan and identify spurious responses.
6. Evaluate zero-IF, low-IF and superheterodyne architectures.
7. Apply practical methods for biasing, matching, filtering, calibration and measurement.
8. Assess contemporary mixer technologies and their limitations.

1. Fundamental Principle of Frequency Conversion

An ideal multiplier produces

where v₁(t) and v₂(t) are input voltages, v_o(t) is the output voltage, and k is a dimensional scaling constant determined by the circuit. If the inputs are sinusoidal,

then, using the product identity,

Thus, multiplication creates the sum and difference frequencies. The units are consistent because k must have units of inverse voltage when all quantities are voltages. In an ideal mixer, one input is called the radio-frequency (RF) or signal port and the other is the local-oscillator (LO) port. The desired output is commonly the intermediate-frequency (IF) component:

The alternative sum component is removed by a filter or is used intentionally in some up-conversion systems.

A real transistor or diode is not an ideal multiplier. Its nonlinear current-voltage characteristic can be expanded around a bias point:

The coefficients depend on bias and device technology. If v = V_RF cos(ω_RF t) + V_LO cos(ω_LO t), the square and higher-order terms generate harmonics and intermodulation products. The second-order term generates ω_RF ± ω_LO, while third-order terms can generate products such as 2ω_RF − ω_LO and 2ω_LO − ω_RF. The Taylor expansion is a local approximation; it is useful for weak-signal analysis but may be inaccurate when the LO drives the device deeply into switching operation.

Frequency-domain interpretation

Multiplication in time corresponds to convolution in frequency. If x(t) is multiplied by p(t), then its spectrum is shifted according to the spectral components of p(t). For a sinusoidal LO, the spectrum contains impulses at +f_LO and −f_LO, producing two translated copies of the input spectrum. This interpretation explains why a mixer translates bandwidth as well as centre frequency.

Figure: A mixer translates an RF spectrum by the LO frequency while generating additional products.

2. Mixers and Their Performance Measures

2.1 Passive and active mixers

A passive mixer, often based on Schottky diodes or MOSFET switching, has no power gain. Its performance is normally stated as conversion loss:

where P_RF is the available RF input power and P_IF is the available power in the desired IF output, measured under specified impedances and frequencies. A positive L_C denotes loss. An active mixer uses biased transistors to provide conversion gain:

Conversion gain and conversion loss must not be compared without specifying port impedances, LO power, frequency, bandwidth and whether powers are available or delivered.

Passive mixers generally offer good linearity, high input compression point, low flicker-noise contribution and broad bandwidth, but require substantial LO drive and add conversion loss. Active mixers can provide gain and require less LO power, but consume DC power and may exhibit poorer linearity, LO-to-RF leakage and additional noise.

2.2 Unbalanced, single-balanced and double-balanced mixers

An unbalanced mixer has limited rejection of LO and RF feedthrough. A single-balanced mixer cancels one class of unwanted signal by symmetry. A double-balanced mixer suppresses both even-order feedthrough components at its output and normally provides useful port isolation. The cancellation is imperfect because of device, layout and transformer mismatch.

The classical diode-ring mixer is a double-balanced passive mixer. Four diodes are driven as a commutating bridge. An adequately large LO approximates a square-wave polarity reversal, so the RF signal is alternately passed with positive and negative polarity. The periodic switching waveform has Fourier components at odd LO harmonics, which explains both desired mixing and harmonic spurs.

2.3 Switching-function model

For an ideal polarity-reversing mixer, the output can be approximated as

The square wave has fundamental amplitude 4/π in normalized form. Consequently, the wanted fundamental mixing coefficient is approximately 2/π for voltage conversion in an ideal switching model, before accounting for impedance and power definitions. This is a model result, not a universal specification: finite LO amplitude, diode resistance, parasitic capacitance, waveform distortion and filtering alter the measured conversion loss.

2.4 Isolation and leakage

Port isolation is the ratio of power launched into one port to undesired power appearing at another port, expressed in decibels. For example,

Higher isolation is desirable. LO leakage can desensitize a receiver, create radiation from an antenna, or produce DC offsets in a direct-conversion receiver. Differential signalling, balanced structures, shielding, careful grounding and calibration improve isolation.

2.5 Noise and linearity

A mixer contributes noise through thermal device noise, LO noise conversion, flicker noise and noise associated with conversion loss. For a passive mixer operated at a common reference temperature, conversion loss establishes a lower bound on noise figure, although practical noise figure can be higher. Noise figure is

where signal-to-noise ratios are linear quantities measured at the same reference bandwidth. In decibels, NF = 10 log₁0 F.

Mixer linearity is commonly characterized by the 1-dB compression point and third-order intercept point (IP3). Two equal interferers at f₁ and f₂ generate third-order products at 2f₁ − f₂ and 2f₂ − f₁. If the fundamental and third-order product powers are measured in dBm at an extrapolated intercept, then, under the usual small-signal straight-line approximation,

where P_fund and P_IM3 are output powers in dBm. This equation is an extrapolation, not a physical operating point; compression and noise invalidate it near large signals.

3. Modulation as Controlled Frequency Translation

Modulation maps information onto a carrier. In amplitude modulation (AM), a real message signal m(t) multiplies a carrier:

where A_c is carrier amplitude, f_c is carrier frequency, m_n(t) is a normalized message satisfying maximum magnitude no greater than one, and μ is the modulation index. For conventional AM without envelope crossing, 0 ≤ μ ≤ 1. The spectrum contains a carrier at f_c and upper and lower sidebands at frequencies displaced by the message frequencies.

For a single-tone message m_n(t)=cos(2πf_mt),

In suppressed-carrier double-sideband modulation (DSB-SC),

The carrier is absent, improving power efficiency but requiring coherent demodulation. Single-sideband (SSB) modulation removes one sideband as well as the carrier, reducing occupied bandwidth and transmitted power. Quadrature methods, phasing networks and digital signal processing can generate SSB.

IQ modulation

Complex-envelope notation is particularly useful in modern transceivers. Let

where x(t)=I(t)+jQ(t) is the complex baseband envelope. In an ideal quadrature modulator, I(t) and Q(t) drive mixers with LO signals separated by 90°. Amplitude and phase imbalance, LO quadrature error and DC offsets cause image leakage and carrier leakage.

A useful image-rejection measure is

or IRR_dB=10 log₁0(IRR). For amplitude mismatch ε and phase error φ in radians, a commonly used small-error approximation is

provided both errors are small and expressed as dimensionless quantities. This approximation should not be used for large mismatches; exact phasor analysis or measurement is then required.

4. Frequency-Conversion Architectures

4.1 Superheterodyne receiver

A superheterodyne receiver mixes the received RF signal to an IF. If the desired RF frequency is f_RF and the LO is above the RF, the image frequency is

For low-side injection, the corresponding image is f_RF − 2f_IF, provided the frequency is positive. Both signals produce the same absolute IF because

The image-rejection requirement therefore places a demand on the RF preselector. For an ideal tuned single-resonator response, an approximate image-rejection ratio is

where Q is the loaded quality factor. This expression assumes a narrowband resonator and small loss; it is a design estimate rather than a substitute for a complete filter response.

Advantages of the superheterodyne architecture include excellent selectivity, mature filtering and flexible gain distribution. Disadvantages include image responses, multiple spurious responses, LO radiation and the need for frequency planning.

4.2 Zero-IF and low-IF receivers

A zero-IF receiver mixes the desired RF channel directly to baseband using quadrature paths. It avoids a separate IF filter and is well suited to integrated circuits and software-defined radios. However, DC offsets, flicker noise, even-order distortion, LO self-mixing and I/Q imbalance are significant concerns.

A low-IF receiver uses a small non-zero IF, reducing DC-offset sensitivity while retaining substantial integration. Its main difficulty is image rejection: the image lies close to the desired channel and must be cancelled by accurate quadrature processing.

4.3 Direct RF sampling and digital conversion

High-speed data converters can sample RF directly or after modest analogue conditioning. Digital down-conversion then uses numerically controlled oscillators, complex multiplication and decimation. The analogue mixer may be eliminated, but sampling creates its own frequency translations. For a sampling frequency f_s, a continuous-time sinusoid at f_in aliases to a frequency related to |f_in − nf_s|, where integer n is selected to place the result in the first Nyquist zone. Anti-alias filtering and clock-jitter analysis remain essential.

5. Spurious Responses and Frequency Planning

A practical mixer generates products represented by

where m and n are integers, including zero where appropriate. The order of a product is often defined as |m|+|n|, though conventions vary. A spur map lists products that fall inside the receive or transmit band and ranks them by expected amplitude.

A disciplined frequency plan should:

• choose an IF that balances image rejection, filter feasibility and oscillator complexity;
• identify LO harmonics and RF harmonics;
• check products against sensitive channels and clock frequencies;
• specify filtering at every conversion stage;
• include frequency tolerances, phase noise and Doppler or reference error;
• evaluate blocker-induced intermodulation.

Worked example: selecting a high-side LO

Suppose a receiver must accept a channel at f_RF=2.450 GHz and use f_IF=150 MHz. With high-side injection,

The image is

Thus, the RF preselector must pass 2.450 GHz while sufficiently attenuating 2.750 GHz. A blocker at 2.750 GHz would otherwise be translated to the same 150-MHz IF. This calculation assumes ideal frequency conversion and ignores oscillator tolerance; practical specifications must add margins for filter skirts and frequency error.

6. Practical Design Methodology

A mixer design begins with a system specification rather than a transistor schematic. Required quantities include RF and LO frequency ranges, IF bandwidth, input power range, conversion gain or loss, noise figure, compression point, IP3, isolation, supply voltage, temperature range and allowable LO power.

Device and topology selection

Diode-ring mixers are attractive for high dynamic range and broadband operation. Gilbert-cell mixers are common in integrated receivers because differential transconductors and switching pairs provide gain and compatibility with active filters. Passive CMOS commutating mixers can achieve good linearity and low flicker noise, but clock drive and switch resistance must be considered. Harmonic-rejection mixers use multiphase LO waveforms and weighted paths to cancel selected harmonics.

Matching and bias

A mixer is a multiport network. Matching one port can alter conversion gain, noise and isolation at another. S-parameters measured under the correct LO state are more informative than small-signal measurements with the LO disabled. Bias networks should prevent LO and RF signals from coupling through supply rails. Differential symmetry, short return paths, controlled impedance and electromagnetic shielding are often more important than nominal component values.

Filtering and LO generation

The LO must have adequate output power, low phase noise and low spurious content. Phase noise is translated by the mixer, so close-in LO noise can become reciprocal-mixing noise when strong blockers are present. PLL reference spurs, fractional-N spurs and supply modulation should be included in the analysis. Output filters must account for insertion loss, power handling, group delay and impedance transformation.

Measurement

A professional mixer characterization normally includes:

Use calibrated signal generators, attenuators and filters. Prevent analyser overload by checking every unwanted tone, not only the desired IF. Report frequency, power, impedance, bandwidth, LO level, temperature and calibration uncertainty.

7. Applied Case: Illustrative Dual-Conversion Receiver

Background

Consider an illustrative 2.4-GHz wireless receiver requiring a 1-MHz channel bandwidth and strong rejection of an out-of-band blocker. A single conversion to a low IF would make the image filter difficult, while direct conversion would create a demanding DC-offset and flicker-noise problem.

Problem

The design must provide high selectivity, tolerate a strong blocker, and remain implementable with integrated RF circuitry and a practical external filter.

Analysis

A first IF of 400 MHz permits a relatively selective RF filter and places the image 800 MHz away from the desired channel for high-side injection. A second conversion to a 10.7-MHz IF allows a mature narrowband ceramic or integrated filter. The first mixer therefore uses an LO near 2.8 GHz for a 2.4-GHz RF input, while the second LO is selected near 389.3 MHz or 410.7 MHz depending on injection choice.

The frequency plan must test products such as LO harmonics, combinations of the two LOs, and leakage from the first IF into later stages. Gain should be distributed so that the first mixer is not overloaded by the blocker, while the later IF amplifier provides sufficient sensitivity without saturating.

Approach and solution

The proposed chain is: antenna preselector → low-noise amplifier → first double-balanced mixer → 400-MHz band-pass filter → second mixer → 10.7-MHz channel filter → IF amplifier and demodulator. A digitally controlled attenuator before the first mixer protects it during large-signal conditions. Differential IF routing and shielding reduce LO leakage. A calibration mode measures gain, I/Q imbalance and DC offsets at production test.

Outcome

The dual-conversion design trades additional components and power for improved selectivity and blocker tolerance. Its principal residual risks are spurious responses from the two-LO plan, phase noise transferred through both mixers, and gain variation over temperature. These are addressed through a spur table, reference-clock discipline, AGC margin and temperature-aware calibration.

Learning points

The case illustrates that a mixer cannot be selected independently of filters, gain distribution, oscillator architecture and interference environment. A frequency plan is a system design document, not merely an arithmetic exercise.

8. Limitations, Governance and Contemporary Developments

Mixer performance is limited by nonlinearity, parasitic coupling, finite switch speed, LO drive, device breakdown, noise and manufacturing variation. At millimetre-wave frequencies, transmission-line loss, packaging and electromagnetic coupling become dominant design concerns. At very low frequencies, flicker noise and DC offsets dominate.

Regulatory compliance requires control of unwanted emissions, local-oscillator radiation, occupied bandwidth and spurious power. Designers must consult the applicable national spectrum regulations and measurement standards. A frequency plan that works electrically may still fail coexistence requirements. Privacy and security also matter: unintended LO leakage or mixer spurs can reveal operating frequencies or interfere with neighbouring systems.

Contemporary developments include digitally assisted calibration, passive mixer-first receivers, N-path filters, harmonic-rejection architectures, wideband software-defined radios, silicon-germanium and RF-CMOS integration, and machine-learning-assisted calibration. AI is most useful as a calibration and diagnosis aid: it can estimate I/Q imbalance, predict temperature-dependent drift or classify spur signatures. It does not remove the need for physical models, traceable measurements or safety and regulatory review.

Best practice is to combine analytical frequency planning with circuit simulation, electromagnetic co-simulation, corner analysis, prototype measurement and post-layout calibration. Designers should maintain a complete spur inventory and document assumptions behind every specification.

Chapter Summary

Mixers translate spectra by exploiting multiplication or nonlinear switching. The basic products are the sum and difference of RF and LO frequencies, but real devices also generate harmonics, intermodulation products and leakage. Passive mixers offer robust linearity and broad bandwidth at the cost of conversion loss and LO drive; active mixers provide gain but consume power and may have lower dynamic range. Modulators use controlled multiplication to create carrier and sideband structures, while IQ architectures support modern complex-envelope modulation.

Superheterodyne receivers provide strong selectivity but require careful image and spur planning. Zero-IF and low-IF receivers improve integration but introduce DC-offset, flicker-noise and image-rejection challenges. Design quality depends on topology, matching, filtering, LO phase noise, isolation, linearity, calibration and measurement discipline. A successful frequency-conversion design is therefore a coordinated system solution rather than an isolated mixer calculation.

Key Terms

Mixer: A nonlinear or switching circuit that translates frequencies.

Local oscillator (LO): A signal used to perform frequency translation.

Intermediate frequency (IF): A translated frequency used for filtering or amplification.

Conversion gain: Output-to-input power ratio for the desired converted signal.

Conversion loss: Input-to-output power loss in a passive conversion process.

Balanced mixer: A mixer using symmetry to suppress selected unwanted products.

Double-balanced mixer: A topology providing suppression of both major feedthrough paths.

Image frequency: An undesired frequency that converts to the same IF as the desired signal.

Spur: An unwanted spectral component generated by mixing, harmonics or interference.

Isolation: Suppression of signal leakage between mixer ports.

Noise figure: Degradation of signal-to-noise ratio caused by a circuit.

IP3: Extrapolated third-order intercept point used to characterize linearity.

Compression point: Input or output level at which gain departs from its small-signal value.

Modulation index: Ratio describing the extent of carrier modulation.

DSB-SC: Double-sideband suppressed-carrier modulation.

SSB: Single-sideband modulation.

IQ modulation: Complex modulation using in-phase and quadrature components.

Image rejection ratio: Ratio of desired-signal power to image power.

Reciprocal mixing: Conversion of oscillator phase noise into in-band noise by a blocker.

Frequency plan: A documented allocation of RF, LO, IF and spur frequencies.

Review Questions

Short Answer Questions

1. Why does multiplication of two sinusoids generate sum and difference frequencies?
2. Distinguish conversion gain from conversion loss.
3. What is the function of a double-balanced mixer?
4. Why is an image-rejection filter required in a superheterodyne receiver?
5. Name two important impairments in a zero-IF receiver.

Long Answer Questions

1. Derive the output spectrum of a sinusoidal multiplier and explain the assumptions involved.
2. Compare passive diode-ring, Gilbert-cell and passive CMOS mixers in terms of gain, linearity, noise and LO drive.
3. Explain AM, DSB-SC and SSB using time-domain equations and spectral reasoning.
4. Describe a complete procedure for developing and validating a two-stage frequency plan.
5. Discuss the advantages and limitations of superheterodyne, zero-IF and low-IF receiver architectures.

Analytical / Application Questions

1. A 915-MHz signal is converted to a 70-MHz IF using high-side LO injection. Calculate the LO frequency and image frequency, and explain the required preselector action.
2. Two equal tones at 1.000 GHz and 1.002 GHz are applied to a nonlinear mixer. Identify the close-in third-order products and explain why they are difficult to filter.
3. An IQ modulator has a small amplitude mismatch and quadrature phase error. Explain how these errors create an image and propose two calibration methods.

Multiple Choice Questions with Answers

Question 1

For an ideal multiplier driven by frequencies f_RF and f_LO, which desired frequency is commonly selected as the IF?

A. f_RF + f_LO

B. |f_RF − f_LO|

C. 2f_RF + f_LO

D. f_RF/f_LO

Correct Answer: B. |f_RF − f_LO|

Question 2

Which mixer characteristic is normally expressed as a positive dB loss for a passive mixer?

A. Conversion loss

B. Noise temperature

C. Isolation gain

D. Modulation index

Correct Answer: A. Conversion loss

Question 3

A receiver uses high-side injection with f_RF = 2.450 GHz and f_IF = 0.150 GHz. What is f_LO?

A. 2.300 GHz

B. 2.450 GHz

C. 2.600 GHz

D. 2.750 GHz

Correct Answer: C. 2.600 GHz

Question 4

For the receiver in the preceding question, what is the high-side image frequency?

A. 2.150 GHz

B. 2.300 GHz

C. 2.600 GHz

D. 2.750 GHz

Correct Answer: D. 2.750 GHz

Question 5

Which architecture is especially associated with DC offset and flicker-noise problems?

A. Zero-IF

B. Traditional high-IF superheterodyne only

C. Passive band-pass filter

D. Antenna matching network

Correct Answer: A. Zero-IF

Question 6

The third-order intermodulation products of two tones at f1 and f2 include:

A. f1 + f2 only

B. 2f1 − f2 and 2f2 − f1

C. f1/2 and f2/2

D. f1 − 2f2 only in all cases

Correct Answer: B. 2f1 − f2 and 2f2 − f1

Question 7

In conventional AM, the modulation index must generally satisfy which condition to avoid envelope overmodulation?

A. μ < 0 only

B. μ ≥ 10

C. 0 ≤ μ ≤ 1

D. μ = π/2 exactly

Correct Answer: C. 0 ≤ μ ≤ 1

Question 8

What is a principal benefit of a double-balanced mixer?

A. It eliminates the need for an LO

B. It provides perfect conversion at every frequency

C. It suppresses selected feedthrough components through symmetry

D. It always provides unlimited power gain

Correct Answer: C. It suppresses selected feedthrough components through symmetry

Question 9

Reciprocal mixing refers to:

A. Conversion of oscillator phase noise into noise near a desired signal

B. Conversion of DC into an ideal carrier

C. Perfect cancellation of all image frequencies

D. Use of two identical antennas

Correct Answer: A. Conversion of oscillator phase noise into noise near a desired signal

Question 10

If the output fundamental power is 0 dBm and the third-order product power is −40 dBm, what is the approximate OIP3?

A. 10 dBm

B. 20 dBm

C. 40 dBm

D. 80 dBm

Correct Answer: B. 20 dBm